Re: [RFC V4 PATCH 7/8] vhost: packed ring support

[Jason Wang]

On 2018å05æ23æ 15:17, Wei Xu wrote:
> On Wed, May 23, 2018 at 09:39:28AM +0800, Jason Wang wrote:
> > On 2018å05æ23æ 00:54, Wei Xu wrote:
> > > On Wed, May 16, 2018 at 08:32:20PM +0800, Jason Wang wrote:
> > > > Signed-off-by: Jason Wang <jasowang@xxxxxxxxxx>
> > > > ---
> > > >   drivers/vhost/net.c   |   3 +-
> > > >   drivers/vhost/vhost.c | 539 ++++++++++++++++++++++++++++++++++++++++++++++----
> > > >   drivers/vhost/vhost.h |   8 +-
> > > >   3 files changed, 513 insertions(+), 37 deletions(-)
> > > >
> > > > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> > > > index 8304c30..f2a0f5b 100644
> > > > --- a/drivers/vhost/vhost.c
> > > > +++ b/drivers/vhost/vhost.c
> > > > @@ -1358,6 +1382,8 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg
> > > >   			break;
> > > >   		}
> > > >   		vq->last_avail_idx = s.num;
> > > > +		if (vhost_has_feature(vq, VIRTIO_F_RING_PACKED))
> > > > +			vq->avail_wrap_counter = s.num >> 31;
> > > >   		/* Forget the cached index value. */
> > > >   		vq->avail_idx = vq->last_avail_idx;
> > > >   		break;
> > > > @@ -1366,6 +1392,8 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg
> > > >   		s.num = vq->last_avail_idx;
> > > >   		if (copy_to_user(argp, &s, sizeof s))
> > > >   			r = -EFAULT;
> > > > +		if (vhost_has_feature(vq, VIRTIO_F_RING_PACKED))
> > > > +			s.num |= vq->avail_wrap_counter << 31;
> > > >   		break;
> > > >   	case VHOST_SET_VRING_ADDR:
> > > >   		if (copy_from_user(&a, argp, sizeof a)) {
> > > 'last_used_idx' also needs to be saved/restored here.
> > >
> > > I have figured out the root cause of broken device after reloading
> > > 'virtio-net' module, all indices have been reset for a reloading but
> > > 'last_used_idx' is not properly reset in this case. This confuses
> > > handle_rx()/tx().
> > >
> > > Wei
> > Good catch, so we probably need a new ioctl to sync between qemu and vhost.
> >
> > Something like VHOST_SET/GET_USED_BASE.
> Sure, or can we expand 'vhost_vring_state' to keep them done in a bunch?

It's port of uapi, so we can't.

Thanks

> > Thanks